Random Walks with Distortion Memory: Projective Dynamics in Polytopes
Date(s) : 24/11/2026 iCal
14h30 - 15h30
Multiplying non-negative integer matrices we consider a dynamical system on projective convex compact polytopes. This kind of systems appears in the Gauss map for continuous fractions and later in the RAUZY induction for interval exchange transformations. The system is conjugated to a random walk on a finite graph: a matrix is attached to each edge and products of matrices are computed along the walk. Here the probability of walking on an edge is conditionned by the past. And the conditionning amount to recording a distortion of the projective polytope. The main phenomenon is that the product of matrices balances the vectors. We state natural hypotheses on the matrices plus a combinatorial condition on the graph that imply ergodicity of the system. The conditions are algorithmic (and already coded). This work is an extension of the work of Charles FOUGERON on win-loose projective systems.
This is a joint work with Charles FOUGERON and Paul MERCAT.
I first swimmed in Logic, then I hiked the hills of Geometric Group Theory before climbing the cliffs of Dynamical Systems: I thank the audience for praising inter-cultural relationships.
Emplacement
I2M Saint-Charles - Salle de séminaire
Catégories



